We examine the probability of an epidemic (the complement of the probability of disease extinction) in a stochastic host-vector model for dengue transmission. The host population is age-structured. Exposure to the vector population is weighted according to age. Initially we consider the introduction of a single infected individual into an entirely susceptible population. We use matrix methods, together with simulation, to calculate the epidemic probability.

Initially we consider the model without seasonal variation. The way in which exposure to the vector population varies with host age has a strong impact on the epidemic probability associated with an infected host, but it has only a weak impact on the epidemic probability associated with an infected vector. In the model with seasonal variation, the probability of an epidemic is additionally dependent on the time at which an infected individual enters the population. We go on to extend the model to include two serotypes and consider the epidemic probability of when a single individual infected with serotype Y is introduced into a population in which serotype X is endemic.

We conclude that understanding how age-structured and time-variable transmission affects the probability of an epidemic can help to identify targeted intervention strategies. Introducing simple structured control strategies that account for demographic or behavioural structures in the population, and vary over the course of the year, could lead to more efficient and effective epidemic prevention.